What are tokens in AI? How many words is 1,000 tokens?

Tokens are the units of text that AI language models read and write. A token can be a whole word, part of a word, a number or a punctuation mark. Before a model like ChatGPT, Claude or Gemini sees your prompt, a tokenizer splits it into tokens and turns each one into a number. The model then predicts the reply one token at a time.

As a rule of thumb from OpenAI's own guidance, one token is about four characters of English text, or three quarters of a word. So 1,000 tokens is about 750 words. The exact count depends on the model and the language.

Searching for "AI tokens" also brings up crypto coins linked to AI projects. That is a different thing entirely. This guide is about tokens in language models.

How tokenization works

Models do not read letters or whole words. They use a fixed vocabulary of tens of thousands of text pieces learned from training data. Common words are usually one token. Rarer or longer words are split into several familiar pieces.

• "cat" → 1 token
• "Trendifico" → several tokens, because the model has never seen the brand name as one piece
• "2026" → may be one or two tokens depending on the tokenizer
• A space and punctuation often attach to the next word or count on their own

This is why token counts are not a simple word count. Code, URLs, numbers and unusual names use more tokens per word than plain English.

Tokens to words: a conversion table

Using the common estimate of 0.75 words per token for English prose:

Google's Gemini documentation gives a similar estimate: about 4 characters per token, with 100 tokens equal to roughly 60 to 80 English words. Other languages often need more tokens for the same meaning, so a Hindi, Arabic or Japanese text can cost noticeably more than its English version.

Why tokens matter

1. Context window

Every model has a context window: the maximum number of tokens it can handle in one request, counting your prompt, any documents or chat history you include, and its reply. When a conversation exceeds the window, older text is dropped or the request fails. Large windows now range into the hundreds of thousands or even a million tokens, but more context also means higher cost and slower replies.

2. Cost

API providers charge per token, usually quoted per million tokens. Input tokens (what you send) and output tokens (what the model writes) are priced separately, and output is typically several times more expensive.

Example at illustrative prices of $1 per million input tokens and $5 per million output tokens: a support bot handles 10,000 chats a month, each with 2,000 input and 500 output tokens. That is 20 million input tokens ($20) plus 5 million output tokens ($25), or $45 a month. Plug in real prices with the AI API Cost Calculator, and read what it costs to add AI to your product for a fuller budget.

3. Output limits and speed

Models also cap how many tokens they will write in one reply. Long answers take longer because each token is generated in turn. Reasoning models add hidden "thinking" tokens before they answer; these count as output and are billed even though you may not see them.

Do images and files use tokens?

Yes. Multimodal models convert images, PDFs and audio into tokens too. A single image can use hundreds or thousands of tokens depending on its size and the model, so sending large screenshots in every request adds up quickly.

How to use fewer tokens

1. Cut the boilerplate. Remove repeated instructions and unneeded examples from system prompts.
2. Send only relevant context. Retrieve the few paragraphs that answer the question instead of pasting a whole manual.
3. Ask for the format you need. "Answer in three bullet points" produces fewer output tokens than an open request.
4. Summarise long chats. Replace old turns with a short summary once a conversation gets long.
5. Use smaller models for simple jobs. Classification and extraction rarely need the largest model.
6. Use prompt caching where your provider offers it, so repeated context is billed at a lower rate.

Clear prompts also save tokens on retries. Our guide to writing AI prompts that get better answers and the AI Prompt Generator help you get it right first time.

FAQ

How many words is 1,000 tokens?

About 750 English words. The exact number depends on the model's tokenizer and on how many long or unusual words the text contains.

Is a token the same as a word?

No. Short common words are often one token, but long or rare words are split into several. On average English text has about 1.3 tokens per word.

Do spaces and punctuation count as tokens?

Usually they do, though a space is often merged into the word that follows it. Heavy formatting, tables and code all increase the token count.

Why do different AI models give different token counts?

Each model family uses its own tokenizer and vocabulary. The same paragraph can be a few percent longer or shorter in tokens on another model, so always check with the provider's own counter before relying on exact numbers.